Windowed UID actions T-Test for Anomaly detection from Nominal Groups
One of the concerns of using density-based methods for Outlier Detection is that a group of adversaries could disturb your data by slowly or all at once creating a dense region in your graph. WUTANG was created to identify when there's a group of activity that is making movements in any feature space that could potentially poison the model.
(Will be filling in this page soon, in the meantime notebook is available)